Determinants of event type in Persian complex predicateshharley/PDFs/HarleyFolliKarimi2003.pdf · Determinants of event type in Persian complex predicates 3 In this paper, we show

„ 2003 by Raffaella Folli, Heidi Harley & Simin KarimiLuisa Astruc & Marc Richards (eds.)

Cambridge Occasional Papers in Linguistics 1:100-120.

Determinants of event type in Persian complex predicates

Raffaella Folli, Heidi Harley & Simin Karimi

University of Cambridge & University of Arizona

In this paper we analyse the interdependence of Persian nonverbal (NV)element and the light verb (LV) in determining the syntactic properties, theevent structure, and the alternation possibilities of the entire complexpredicate (CP) and we argue that these properties provide strong evidenceagainst a Lexicalist approach to such phenomena. Further we show howthese facts may be naturally accommodated within a constructionalistposition such as Hale and Keyser (1993, 2002), and argue that thecombination of compositionality and syntactic independence effectsobserved in these constructions, are difficult, if not impossible, to deal within a projectionist approach.

1 INTRODUCTION

It has been argued in the literature that the argument and event structures of Persian complexpredicates (CPr), as well as syntactic properties such as control, cannot be simply derivedfrom the lexical specifications of the nonverbal element (NV) or the light verb (LV), thereforesuggesting that the syntactic and semantic properties of these elements must be determinedpost syntactically rather than in the lexicon (Karimi 1997). In this paper, we show that theevent structure of LV is not always the same as the event structure of its heavy counterpart.Furthermore, although LV determines the agentivity (xordan 'collide' versus zadan 'hit') andthe eventiveness of the CPr, it fails to completely determine the event structure and the telicityof the CPr. Thus, depending on the NV element, the same LV may occur in different types ofevent structure. For example, the LV xordan 'collide' may occur in both accomplishment andachievement complex predicates, while the LV zadan 'hit' can occur in activity ,accomplishment, and semelfactive complex predicates, when combined with different NVelements. We argue that when the LV allows for event type variation (as in the case of xordan'collide'), it is the category of the NV element that determines the event structure of the wholeCPr. That is, if the NV element is a noun, the CPr is atelic (activity or semelfactive), unlessthe noun is eventive (see section 5), in which case the CPr may be telic (accomplishment)). Ifthe NV element is an adjective, an adverbial particle, or a prepositional phrase, the CPr isalways telic (accomplishment or achievement). This is summarized in (1):

(1) Event StructuresCategory of NV Telic AtelicNoun (if not eventive) * √A/Adv Particle/PP √ *

However, there are also cases where the event type of the complex predicate is determined bythe LV alone, and not the NV element. This is the case of shodan ‘become’ which gives riseonly to accomplishments and achievements, due to its inherently telic meaning which doesnot allow for aspectual variation (see section 4.3.3). (This inherently telic meaning may turnout to be reducible to shodan’s selectional properties, if the current proposal is on the righttrack).

Raffaella Folli, Heidi Harley & Simin Karimi2

We go on to show that the semantics of the NV element determines whether it cancombine with particular LVs. Finally, we discuss certain predictions that follow from ouranalysis of Persian CPr.

As it can be inferred by these very preliminary considerations, the interdependenceand systematicity of the NV element and LV’s contributions to determining the eventstructure and alternation possibilities of the entire CPr seem to be strong evidence against aLexicalist approach to such phenomena. Accordingly, in this paper we show how these factsmay be naturally accommodated within a syntax-based approach to argument structure, andargue that the combination of compositionality and syntactic independence effects observed inthese constructions are difficult, if not impossible, to deal with in a projectionist approach.

The traditional GB-style approach to projection involves representing verbs completewith their argument structures in the lexicon, which then project into the syntax. Accordingly,the Projection Principle (Chomsky 1986:84) states that lexical information must besyntactically realized. The argument structures of the verbs are linked via universal principlesto particular syntactic positions.

In such a theory, argument-structure alternations, whether morphologically marked ornot, are accomplished via a separate generative process that occurs within the lexicon, prior toprojection. For instance, a transitive verb may be mapped to an intransitive verb via thelexical rule of Passive, which alters both the argument structure and the morphology of therelevant verb. The altered lexical entry then projects in accordance with the linking principles,thereby indirectly giving rise to the altered syntax of passives. The same kind of explanationhas then been adopted by various lexicalist analyses to explain not only active/passivealternations, but also many other kinds of alternations that verbs display in languages likeEnglish (Levin and Rappaport 1995, among many others).

Beginning with Baker, Johnson and Roberts (BJR) (1989), and realized most fully inthe work of Hale and Keyser (1992, 1993 and subsequent work.), however, a sustained efforthas been made to eliminate lexical rules and generate all argument structure alternations in thesyntax, greatly simplifying the model of the lexicon. In such “constructionalist” theories, theverb is inserted into a particular complex syntactic structure, which determines the locationand interpretation of each of the arguments in the verb phrase. Argument structurealternations then become a matter to be treated in the syntax, rather than in the lexicon. TheBJR treatment of passive, for instance, involved treating the passive morpheme as anargument of the verb, which saturated the verb's external argument position and was thensuffixed to the verb in the syntax. Hale and Keyser's approach is even more radical.Unergative verbs are created by incorporating the object in a transitive structure into anabstract verbal head, which then appears to be intransitive. Work is underlyingly transitive:"do work", as in (2a) below. Argument structure alternations are created when the same rootappears in different syntactic structures (see (2b-c)).

(2) a. vP b. vP c. vP

DP v' v AP DP v'

John v N BECOME DP A John v AP

DO work the door open CAUSE DP A

the door open

"John worked" = "The door opened" = "John opened the door" ="John did work" "The door became open" "John caused the door open"

Determinants of event type in Persian complex predicates 3

In this paper, we show that two of the Hale and Keyser structures above map naturally ontothe Persian CPr constructions, accounting for their varying event structure and agentivity.Evidently, the Persian CPr constructions in many cases look like an obvious one-to-one matchwith the underlying syntactic representations of argument structure, assuming thatincorporation of the NV element into the LV does not take place (and allowing for the head-final nature of Persian).

The article is organized as follows. In section 2 we look at phrase structure of Persianin general and the way the language forms Complex Predicates in particular. We show that anumber of pieces of evidence can be adduced in support of the independent syntactic nature ofthe LV and the NV element. In section 3 we briefly summarize Hale and Keyser’s frameworkfor deriving argument structure and verb alternations. In section 4 we analyze the effect ofeach element of the complex predicates in determining the aspect properties of the whole andwe discuss different types of complex predicates, depending on the categorical nature of theNV element. The phrase structure of eventive NV elements is discussed in section 5. Thecompatibility of the NV element with LV is discussed in section 6. Finally in section 7 welook at some other cases of possible and impossible alternations that our analysis is able topredict. Section 8 concludes this paper.

2 PHRASE STRUCTURE OF PERSIAN

2.1 General Background

Persian is a verb final language that exhibits the following unmarked word order in a doubleobject construction:

(3) a. S Ospecific PP Vb. S PP Ononspecific V

The specific direct object appears in a higher position, preceding the indirect object. Thenonspecific object is adjacent to the verb, following the indirect object. This is a propertyseen in many other languages such as Hindi, Turkish, German, and Dutch. Examplesillustrating (3a) and (3b) are provided in (4a) and (4b), respectively.

(4) a. Kimea ketâb-ha ro be Papar dâdK book-pl râ to Papar gave'Kimea gave the books to Papar.'

b. Kimea be Paper ketâb dâdKimea to Paper book gave'Kimea gave book(s) to Papar.'

(5) exhibits the phrase structure underlying both (4a) and (4b) (Karimi, to appear):

(5) vPS v'

VP vPP V'

DP V+/- specific kard


The surface order in (3a) is obtained by movement of the [+specific] object, which is followedby the specificity marker râ, to the edge of vP. Accusative Case on the object is checked inthat position prior to its movement out of vP in order to escape the existential closure. (Ofcourse, the subject in spec-vP moves further leftward yet, to Spec-TP or higher, in order tocheck its case features; this gives the final S Ospec PP V word order). The nonspecific objectremains in situ, directly generating the word order in (3b).1

2.2 Complex Predicates

2.2.1 OverviewComplex verbs have gradually replaced simple verbs in Persian since the thirteenth century.The tendency to form complex verbs has resulted in the existence of two sets of verbs, simpleand complex, for a number of verbal concepts. In many cases, the application of the simpleverb is restricted to the written and elevated language. A few examples of simple/complexpairs appear in (6) (See Karimi, 1997 for more examples). The productivity of CPr formationis such that it has completely replaced the former morphological rule of simple verb formationin this language (Bateni, 1989).

(6) Simple Complexlasidan las zadan (flirtation doing) 'to flirt'raghsidan raghs kardan (dance doing) 'to dance'agahanidan agah kardan (informed making) 'to inform'aghazidan aghaz kardan (start doing) 'to start'

The LV of Persian CPr ranges over a number of simple verbs, as shown by Karimi (1997). Asample of LVs employed in CPr constructions is provided in (7).

(7) a. kardan 'to do' l. budan 'to be'b. shodan 'to become' m. chidan 'to arrange'c. xordan 'to collide' n. gereftan 'to catch, to take'd. zadan 'to hit' o. keshidan 'to pull'e. dâdan 'to give' p. nemudan 'to show'f. dâshtan 'to have' q. oftâdan 'to fall'g. âmadan 'to come' r. pâshidan 'to scatter'h. andâxtan 'to throw' s. raftan 'to go'i. âvardan 'to bring' t. sepordan 'to entrust'j. bastan 'to tie' u. shostan 'to wash'k. bordan 'to carry' v. gozashtan 'to pass, to cross'

The light verb kardan 'to do/make' has almost entirely lost its heavy interpretation, and is themost productive LV. The LV shodan 'to become' is systematically used in so-called passiveor unaccusative constructions.

Another characteristic of Persian CPr is that its NV element ranges over a number ofphrasal categories, as exemplified by (8) (see Karimi, 1997 for additional examples).

1 Karimi (2003) suggests two distinct underlying object positions: the nonspecific object is base-generated as a sister to the verb, and the specific one in the Spec of VP. The structure in (5) differs from thatproposal in that the specific object and its nonspecific counterpart are both base generated in the same position.The two proposals have one important property in common: the specific object receives its interpretation in itssurface position, that is in a position preceding the indirect object. In the spirit of Baker (1988, 1996), it isassumed that the nonspecific object, being inside the predicate construction, does not need Case. For detailedanalysis see Karimi (to appear).


(8) a. N+LVkotak zadan/xordan (beating hitting/colliding) 'to beat, to get beaten'xar kardan/shodan (donkey doing/becoming) 'to fool, become fooleddust dâshtan (friend having) 'to love'

b. A+LVsabok kardan/shodan (light doing/becoming) 'to degrade' (tr & intr)pahn kardan/shodan (wide doing/becoming) 'to spread, to widen' (tr & intr)derâz keshidan (long pulling) 'to lie down, to take a nap'

c. Particle+LVbirun kardan (out doing) 'to dismiss, to fire (someone)'bâlâ âvardan (up bringing) 'to vomit'bâlâ keshidan (up pulling) 'to steal'

d. PP+Vbe yâd dâshtan (to memory having) 'to have in memory'bejâ âvardan (to place bringing) 'to recognize'be bâd dâdan (to wind giving) 'to waste'

Finally, the NV element of Persian CPr may also be a complex phrasal element, as in (9):

(9) Complex NV elementdast o pâ kardan (hand and foot doing) 'to try (hard)'sar o kâr dâshtan (head and work having) 'to be involved'dast be dast kardan (hand to hand doing) 'to hesitate'

We will not discuss this type of NV element in this paper.

2.2.2 The syntactically independent nature of the LV and the NV element in PersianA Persian CPr cannot be considered a lexical unit since its NV element and LV may beseparated by a number of elements, including (a) negative and inflectional affixes, (b) theauxiliary verb for future tense, and (c) emphatic elements (Mohammad and Karimi 1992).2

Furthermore, the NV element of Persian CPr allows limited modification, as in (10).

(10) a. Kimea az ra'is-e edâre [CV [NV da'vat-e rasmi] kard ]] Kimea of boss-Ez office invitation-Ez formal did 'Kimea extended a formal invitation to the boss of the office.'b. Kimea barâye in xune [CV [NV chune-ye xubi ] zad ]] Kimea for this house chin-Ez good hit 'Kimea performed a good negotiation for this house.'

The adjective rasmi 'formal' modifies the nominal NV element in (10a), while xubi 'good'modifies the NV element chune in (10b).

Gapping is also allowed in the case of Persian CPr:

2 Abbreviations:râ = Specificity Marker for Accusative Case pl = plural sg = singularhab = habitual emph = Emphatic neg = negation Ez = Ezafe particleThe Ezafe construction involves a DP consisting of a head noun (an element with the feature [+N] such as N orA), its modifier(s), an optional possessive DP, and the Ezafe particle e that is structurally utilized as a linkbetween the head and its modifiers (and the possessive DP). For recent analysis of Ezafe constructions seeGhomeshi (1996).


(11) Kimea faghat man-o da'vat karde, to-ro ke __ na-kardeKimea only me-râ invitation did, you-râ emph __ neg-did'Kimea has only invited me, not you.'

Finally, Persian NV elements cannot be scrambled out of V' (Karimi 2003) unless theycontain a quantificational element and receive heavy stress, as attested by the contrast in (12).

(12) a. Kimea [che zamin-e saxti ]i diruz [CV ti xord ]Kimea what earth-Ez hard yesterday collided'What a hard fall Kimea had yesterday.'Lit. Kimea what a hard earth yesterday collided.

b. *Kimea zamin diruz xord Kimea earth yesterday collide

These examples suggest that the LV and the NV element in Persian CPr are separatelygenerated and combined in syntax, and become semantically fused at a different, later level.The two parts of CPr enjoy syntactic freedom to a certain degree; nonetheless, their semanticproperties are the same as those of single word elements elsewhere in Persian and in thegrammars of languages like English. These conflicting properties can be easily accomodatedin a non-Lexicalist theory (like, for instance, Distributed Morphology or other radicalconstructionalist theories like that proposed by Borer 2002), where all interpretation occurspost-syntactically. They pose a more serious problem for lexicalist accounts, which wouldessentially need to claim that Persian Complex Predicates are instances of ‘idioms’, receivinga separate entry in the lexicon complete with their syntactic structure. As noted by Marantz(1997), there is no principled independent way of distinguishing between the meanings of so-called ‘idioms’ and the meanings of single-word elements like ‘cat’ or ‘pacify’.

3 INTRODUCTION TO HALE AND KEYSER’S SYNTACTIC ARGUMENT STRUCTURE

As outlined above, Hale and Keyser (1993 et. seq., esp 2002; henceforth H&K) propose aradical new approach to argument structure. Verbs, even in English, are not syntacticallysimplex items, but rather are composites of a light verb and a non-verbal syntactic element.The surface form of the verb results from incorporation of one or more heads in the non-verbal constituent with the light verb.

Their analysis deals with three main kinds of non-verbal constituent: bare N heads,adjectival small clauses, and prepositional small clauses. Their analysis draws its primaryinspiration from English, where the categorial status of adjectival and nominal verb roots isvery clear. They propose that denominal and deadjectival verbs are derived from threeprimary underlying structures:

(13) a. Deadjectival verbs b. Denominal unergative verbsVP VP

VBECOME AdjP DP V’

-en DP Adj Agent VDO N

Theme red ∅ dance

(e.g. The sky reddened) (e.g. Bill danced)


c. Denominal location/locatum verbs:VP

DP V’

Agent V PP

∅ DP P’

Theme P N

∅ saddle

(e.g. The cowboy saddled the horse)

This approach immediately explains many puzzles, both theoretical and empirical. Amongother things, it makes the difference between unergative and unaccusative verbs depend onmore than the X-bar notation. It explains the morphological properties of English verbs ofthese classes. In many languages, the verbalizing part of the structure is visiblymorphologically realized as an affix, as in these examples from Jemez, taken from H&K1993:

(14) a. sáae'-a b. záae'-a c. se-/awork-do song-do word-do“work” “sing” “speak”

Here, the V portion, so often a zero morpheme in English, is realized as the suffix -a, ‘do’,attached to a clearly nominal element. Even in English, the various V heads are often overtlyrealized; the -en suffix is arguably such a morpheme, as are -ize (as in criminalize), -ify (as inclarify), and -ate (as in marinate).

On such an approach, the thematic properties of a particular verb are dependent on thesyntactic and semantic properties of the verbalizing functional element and of the non-verbalconstituent which make it up. Changing the properties of the verbalizing element — the lightverb — results in a change in Agent selection: the light verb is responsible for the presence orabsence of an external argument. (Hence, on this approach, Passive is naturally seen as theresult of a change in choice of light verb, not as a result of a lexical operation. Similarly, thecausative/inchoative alternation in pairs like John opened the door/The door opened is alsothe result of varying the light verb, although the morphological consequences of this variationare invisible in English).

Harley (2001) argued that the syntactic and semantic properties of the non-verbalconstituent are responsible for the internal event structure of the final composed predicate.Simple N complements, as in the denominal unergative verbs, behave as Incremental Themes,measuring-out the event by virtue of their inherent boundedness properties (hence, e.g. danceis atelic, but foal is telic). Predicative complements, as in the verbs based on adjectival andprepositional non-verbal constituents, function as small-clause results, measuring-out theevent by virtue of the inherent boundedness or lack thereof of their incremental status (hence,e.g. redden is atelic, because a thing can continue to become more intensely red for anarbitrary period, but clean is telic, since once something is clean, it can’t get cleaner —cleanliness is inherently bounded (see Weshcler, 2001, Hay, Kennedy & Levin, 1999, Folliand Harley, 2002)). Finally, of course, the properties of the nonverbal constituent determine


the number of internal arguments present: 0 (as in unergatives), 1 (as in unaccusatives andtransitives) or 2 (as in ditransitives).

Below, we will show that each of H&K’s proposed underlying structures for Englishverbs, above, have natural non-incorporated counterparts in Persian complex predicate (CPr)constructions, where the light verb and non-verbal element are realized separately. Further,we will show that the agentivity of a particular CPr is dependent on the light verb involved,and the telicity of the CPr is dependent on the non-verbal element involved, in a verytransparent fashion. Persian, therefore, is a language in which the complex syntactic nature ofverbs is very easily discerned, and in which Hale and Keyser’s proposals concerning thestructure of the verb phrase find striking confirmation, despite the fact that they wereoriginally designed to account for the facts of a typologically extremely dissimilar language.

4 DETERMINANTS OF EVENT STRUCTURE IN CPR

4.1 Deriving unergative, inchoative, and causative argument structures

In the previous section we saw that unergatives are formed when a nominal element isincorporated into a light verb which selects for an external argument. Similarly, inchoativesresult when an adjectival element is incorporated into a light verb which does not select for anexternal argument. These structures translate naturally to Persian CPr. Consider therepresentation of a CPr like gerye kardan, 'weeping doing', that translates as a typicalunergative like cry:

(15) a. vP b. vP

DP v' DP v'

John v N Kimea N v

DO cry gerye 'cry' kard 'did'

'John cried' 'Kimea cried'

Similarly, consider the syntax of a CPr that translates as a typical inchoative, like bidârshodan 'awake becoming':

(16) a. ‘John awoke’ vP b. ‘Kimea awoke’ vP

v AP AP v

BECOME A DP DP A

awake John Kimea bidâr shod 'awake' 'became'

Just as hypothesized by Hale and Keyser for the English causative/inchoative alternation, thealternation between the inchoative and the causative of awake in Persian is accomplished bychanging the light verb from the equivalent of 'become' (shodan) to the causative 'make'(kardan).


(17) a. vP b. vP

DP v' DP v'

Sue v AP Papar AP v

CAUSE A DP D A

awake John Kimea-ro bidâr kard awake make

It should be clear from the above that the Persian case constitutes the strongest possibleevidence for the syntactic nature of l-syntax.

Above, we have seen that altering the particular light verb in a Persian CPr can affectthe appearance or absence of an Agent argument, as expected on a vP-shell theory ofargument structure. We show below that this is a general property of the LV in the CPr.Further, we demonstrate the tight relationship between event type and the category of the NVelement in the CPr. That is, the category of the complement to v determines the event type ofthe CPr, when the LV itself is not inherently telic.

4.2 What the LV can do

4.2.1 Agentivity and CausativenessThe choice of LV determines whether or not the CPr selects for an agent (Karimi, 1997,Megerdoomian, 2002a). This is shown in the following contrasts.

(18) a. tim-e mâ unâ-ro shekast dâdteam-EZ we they-râ defeat gave'Our team defeated them.'

b. tim-e mâ az unâ shekast xordteam-EZ we of they defeat collided'Our team was defeated by them.' (lit. our team got defeat from them.)

As in the case above, the alternation between an agentive and non-agentive structure isaccomplished by selecting a different light verb; we have moved from a causative to aninchoative argument structure with the shift from agentive dâdan ('give') to inchoative xordan('collide'). A similar pair can be seen in (19) below:

(19) a. Minu bachcha-ro kotak zadMinu child-râ beating hit'Minu hit the child.'

b. bachche kotak xordchild beating collided'The child got hit.'

If we go back to our list of LV in (7), we see that the Agent-selecting properties of any givenlight verb are consistent across all Complex Predicates formed with a given LV. We can showthis because the grammaticality of an agentive adverbial such as amdan 'intentionally' remainsconstant even when the NV element’s category is manipulated. In (20) and (21) below wegive evidence for this with respect to two LVs zadam 'to hit' and xordan 'to collide' (HVstands for Heavy Verb).


(20) zadam 'to hit'a. Kimea amdan bachcha-ro zad HV

Kimea intentionally child-râ hit'Kimea hit the child intentionally.'

b. Kimea amdan be ghazâ dast zad LVK intentionally to food hand hit'Kimea intentionally touched the food.'

c. Kimea amdan dâd zad LVK intentionally yell hit'Kimea yelled intentionally.'

d. Kimea amdan dast zad LVK intentionally hand hit'Kimea clapped intentionally.'

(21) xordan 'to collide'a. *Kimea amdan be divâr xord HV

K intentionally to wall collided'Kimea intentionally hit the wall.'

b. *ghazâ amdan dast xord LVfood intentionally hand collided'Food became intentionally touched.'

c. Kimea amdan shekast xord LVK intentionally defeat collided'Kimea intentionally got defeated.'3

We consider this strong evidence for the contention that Agents are selected for by a differentpredicate than other arguments, cross-linguistically. This has been argued for notably byKratzer (1996) and Marantz (1997), on purely semantic grounds (the unavailability ofidiomatic interpretations of agents+verb, to the exclusion of the object) in languages wherethe complex vP structure is morphologically invisible. Here in Persian, the complex structureis transparent, and it is clear that agentivity is a property of the LV in the CPr, and neverdepends on the nature of the NV element selected.

The only cases where choice of NV element appears to affect the projection of anAgent argument is with verbs of motion, like pass and come, as illustrated in (22) and (23)below.

(22) gozashtan 'to pass'a. Kimea amdan az xiyâbun gozasht HV

K intentionally of street passed'Kimea intentionally crossed the street.'

b. *Kimea amdan dar gozasht LVK intentionally away passed'Kimea intentionally passed away.'

(23) âmadan 'to come'a. Kimea amdan âmad HV

K intentionally came'Kimea intentionally came.'

3 As in English, this is only grammatical on a coercion reading, where the subject agentively did someaction that resulted in his/her purposeful defeat. If we substitute a subject which is incapable of havingintentions, we can see that the result will be ungrammatical:(i) *asb-e sefid amdan shekast xord

horse-Ez white intentionally defeat collided (lit. *The white horse got defeat intentionally)


b. *Kimea amdan be donyâ âmad LVK intentionally to world came'Kimea was born intentionally.'

Verbs of motion in many languages alternate between an agentive/unergative and aninchoative/unaccusative reading: (compare German sentences in: Johann ist nach Hausegefahren 'John went home (by car, someone else drove the car)' and Johann hat nach Hausegefahren 'John drove home'); we consider this alternating behavior to be characteristic ofverbs of motion also in Persian. Similarly, the causativity of CPr is also determined by the LV, as suggested byMegerdoomian (2002b). In (24) and (25) below we consider two examples:

(24) a. âb be jush âmadwater to boil came'The water boiled.'

b. Nimâ âb-ro be jush âvardNima water-râ to boil brought'Nima boiled the water.' (Megerdoomian 2002b)

(25) a. Homa be gerye oftâdHoma to crying fell'Homa started to cry.'

b. Nima Homa-ro be gerye andâxtNima Homa-râ to crying dropped'Nima made Homa (start to) cry.' (Megerdoomian 2002b)

In both cases, the non-verbal element is the same (jush ‘boil’ and gerye ‘crying’), but the CPrchanges from the inchoative âmadan 'to come' in (24) to the causative andâxtan 'tothrow/drop' in (25).

4.2.2 States and eventsIn addition to determining whether the CPr is causative and its external argument is agentive,the light verb distinguishes between eventive and stative CPrs. In the examples below we seethat dashtan is stative (both in its heavy (26) and light form (27)) and therefore it isungrammatical in the progressive form, as typical of statives.

(26) Have as a heavy verba. Kimea ye sag dâr-e

K one dog have-3sg'Kimea has a dog.'

b. *Kimea dâr-e ye sag dâr-eK. have-3sg one dog have-3sgLit. *Kimea is having a dog.

(27) a. Kimea Papar-o dust dâr-eK. P. -râ friend have-3sg'Kimea loves papar.'

b. *Kimea dâr-e Papar-o dust dâr-eK. have-3sg P.-râ friend have-3sgLit. *Kimea is having love Papar.

If we alter the LV while keeping the nonverbal element constant, we see that the stativity ofthe construction changes, suggesting that normally the eventiveness of a complex predicate


depends on the light verb involved and not on the non-verbal element. We can see this in (28)below:

(28) a. Kimea esm-e un-o be yâd dâr-eK. name-Ez her-râ to memory have-3s'Kimea has her name in her memory.'

b. *Kimea esm-e un-o dâr-e be yâd dâr-eK. name-Ez her-râ have-3sg to memory have-3sgLit. *Kimea is having her name in her memory.

c. Kimea esm-e un-o be yâd mi-yar-eK. name-Ez her-râ to memory hab-bring-3sg'Kimea remembers her name.'

d. Kimea esm-e un-o dâr-e be yâd mi-yâr-eKimea name-Ez her-râ have-3sg to memory hab-bring-3sg'Kimea is remembering her name.'

4.2.3 DurationAnother property that depends on the LV is the the duration of the CPr as noted byMegerdoomian 2002a. In (29) the light verb keshidan 'to pull' implies duration of the event,while the the light verb zadan 'to hit' contributes punctuality to the meaning of the complexpredicate. In (30), although both Complex Predicates mean 'to yell', (30b) implies duration.

(29) a. dast zadan (hand hitting) b. dast keshidan (hand pulling) 'to touch'(30) a. dâd zadan (yell hitting) b. dâd keshidan (yell pulling) 'to yell'

4.2.4 SummaryThe following chart summarizes what the LV does within a CPr.

(31)The role of LV in CPr1. Agentivity/Causativity2. Eventiveness3. Duration

4.3 What the NV element can do

In this section we discuss the role of the NV element and its contribution to the aspectualinterpretation of the whole CPr. An overview is presented in 4.3.1, followed by data in 4.3.2.The summary of this section is provided in 4.3.3.

4.3.1 OverviewIn a constructionalist system like Hale and Keyser’s, there is a correspondence between thetype of embedded structure below the vP and the Aktionsart of the whole predicate. Considerthe structures for unergatives, inchoatives and causatives above, repeated here:


(32) a. vP b. vP c. vP

DP v' v AP DP v'

John v N BECOME DP A John v AP

DO work the door open CAUSE DP A

the door open

"John worked" = "The door opened" = "John opened the door" ="John did work" "The door became open" "John caused the door open"

The unergative predicate is characteristically an Activity, in Vendlerian terms, while theinchoative and causative are Accomplishments. We can see this using the standard tests forevent structure below:

(33) a. John worked for 3 hours /#in 3 hoursb. The door opened #for 3 minutes4/ in 3 minutesc. John opened the door #for 3 minutes/ in 3 minutes

(34) a. John is working \ John has workedb. The door is opening ~\ The door has openedc. John is opening the door ~\ John has opened the door

The crucial difference between the two classes seems to be the type of clause that appears inthe complement of v: When the verb denotes a telic Accomplishment, the lower phrase is apredicate and its subject — a small clause indicating a change of state. When the wholepredicate denotes an Activity, the lower phrase incorporating into the verbal shell is a nominalexpression.

Turning to Persian, let’s consider the contrast between ‘bidar shodan’ awake (intr) and‘bidar kard’ awake (tr), illustrated below. In the alternation between the causative and theinchoative form the LV changes from kardan to shodan, but the Aktionsart is not affected,because the complement of the LV is an adjectival small clause in both cases. In contrast, thesame LV kardan is used in awake (tr) and cry (unergative), and yet the Aktionsart of the twoconstructions is different, as we can see using the tests below:

(35) a. Kimea ye sâ'ate/* barâye ye sâ'at bidâr shodK. one hour/for one hour awake became'Kimea became awake within an hour.'

b. Kimea ye sâ'ate/ *barâye ye sâ'at Papar-ro bidâr kardK. one hour/for one hour P.-râ awake did'Kimea woke Papar up within an hour.'

c. Kimea *ye sâ'ate/ barâye ye sâ'at gerye kardK. one hour/for one hour cry did'Kimea cried for one hour.'

4 Of course, in these constructions and in their Persian counterpart, there is a grammatical reading of 'foran hour' that modifies the result state that is syntactically represented by the adjectival phrase. Theungrammatical reading is one in which the actual event of becoming open goes on for an hour. The result-modification reading of 'for an hour' is in fact predicted on the syntactic decomposition approach, as the PP mayadjoin directly to the AP [door open], and express the length of time that the open state lasted.


The same picture is true of cases where the small clause contains a prepositional, rather thanadjectival, NV element. The preposition functions as the predicate of the small clause whichintroduces a result to the event structure of the CPr as a whole. Above, we illustrated thestructures we assume for adjectival and nominal complements to LVs. We can extend Haleand Keyser's account of denominal location/locatum verbs to CPrs with a prepositional NVelement, which will contain a small clause complement to vP, exactly as the adjectival onesdo. The only distinction is that the predicate, rather than being adjectival, is prepositional.

(36) "John shelved a book" (H&K 1993)vP

DP v'

John v PP

CAUSE DP P'

a book P N

(on) shelf

Again, for NV elements that are PPs, the H&K structure will translate directly:

(37) be donyâ âmadan (to world coming) 'to be born'

vP

PP v

DP P' âmad 'came'

Kimea P N

be 'to' donyâ 'world'

Similarly, in cases where a particle, rather than a full PP, is the NV-element, the samestructure will apply:

(38) bâlâ bordan (up carrying) 'to promote'vP

DP v'

Kimea PP v

DP P' bord 'carried'

Papar-o P

bâlâ 'up'


In these cases, as for the adjectival SC cases above, it is the presence of the downstairspredication that is responsible for the telic interpretation of CPr.

4.3.2 DataThe dependence of the Aktionsart on the NV element but not on the light verb is even clearerwhen we consider the data below. The following tests, using temporal adverbials sensitive totelicity, examine different LVs when used as main verbs ('heavy' verbs), and compared withtheir light counterparts. For each LV, different types of NV elements are employed, and wecan see in each case that a change in the category of the NV element determines a change inthe event structure of the complex predicate. The data is summarized in the table in (53).

(39) HV âmadan 'to come'a. Kimea âmad.

K came.'Kimea came.'

b. *Kimea kâmelan âmadK completely came

c. Kimea *barâye ye sâ'at/ ye sâ'ate be kelâs âmadK for one hour/in an hour to class came'Kimea came to class for one/in one hour hour.' OK as 'she spent one hour inclass.'

d. Kimea dâr-e be kelas mi-yâ-dKimea have-3sg to class hab-come-3sg'Kimea is coming to class.'

(40) PP + LV be donyâ âmadan (to world coming) 'to be born.'a. Kimea diruz be donyâ âmad.

K yesterday to world came.'Kimea was born yesterday.'

b. Kimea *kâmelan/*barâye ye sâ'at/? ye sâ'ate be donyâ âmadK completely/for an hour/within one hour to world came'Kimea was born within one hour.'

c. Kimea dâr-e be donyâ mi-yâ-dKimea have-3sg to world hab-come-3sg'Kimea is about to be born.'

(41) HV zadan 'to hit'a. Minu ?kâmelan/ barâye ye sâ'at/*ye sâ'ate Papar-o zad

M completely/for an hour/within one hour P-râ hit'Minu hit Papar for an hour.'

b. Minu dâr-e Papar-o mi-zan-eM. have-3sg P-râ hab-hit-3sg'Minu is hitting Papar.'

(42) N + LV dast+zadan (hand-hitting) 'to touch'a. Kimea *kâmelan/ *barâye ye sâ'at/*ye sâ'ate be ghazâ dast zad

K completely/for an hour/within an hour to food hand hitb. Kimea dâr-e be ghazâ dast mi-zan-e

K have-3sg to food hand hab-hit-3sg'Kimea is touching the food.'

(43) HV xordan 'to collide'a. Kimea kâmelan/ *barâye ye sâ'at/*ye sâ'ate be divâr xord

K. completely/for an hour/within an hour to wall collided'Kimea completely hit the wall.'


b. Kimea dâr-e be divâr mi-xor-eK have-3sg to wall hab-collide-3sg'Kimea is about to hit the wall.'

(44) N + LV shekast xordan (defeat colliding) 'to be defeated'a. Kimea kâmelan/ *barâye ye sâ'at/ye sâ'ate shekast xord

K. completely/for an hour/within an hour defeat collided'Kimea got completely/within an hour defeated.'

b. Kimea dâr-e shekast mi-xor-eK have-3sg defeat hab-collide-3sg'Kimea is about to get defeated.'

(45) HV dâdan 'to give'a. Kimea *kâmelan/*barâye ye sâ'at/*ye sâ'ate ketâb-ro be Papar dâd

K completely/for and hour/within an hour book-râ to P. gaveb. Kimea dâr-e ketâb-ro be Papar mi-d-e

K have-3sg book-râ to P. hab-give-3sg'Kimea is giving the book to Papar.'

(46) N + LV shekast dâdan (defeat giving) 'to defeat'a. Kimea kâmelan/ *barâye ye sâ'at/ye sâ'ate Papar-o shekast dâd

K completely/for an hour/within an hour P.-râ defeat gave'Kimea defeated Papar completely/within an hour.'

b. Kimea dâr-e Papar-o shekast mi-d-eK have-3sg P.-râ defeat hab-give-3sg'Kimea is defeating Papar.'

(47) HV andâxtan 'to throw'a. Kimea *kâmelan/ *barâye ye sâ'at/*ye sâ'ate gol-ro andâxt

K. completely/for an hour/in an hour flower-râ threwb. Kimea dâr-e gol-ro mi-y-andâz-e

K have-3sg flower-râ hab-throw-3sg'Kimea is about to throw the flower.'

(48) N + LV dast andâxtan (hand throwing) 'to mock'a. Kimea kâmelan/ barâye ye sâ'at/*ye sâ'ate Papar-o dast andâxt

K. completely/for an hour/within an hour P.-râ hand threw'Kimea completely/for an hour mocked Papar.'

b. Kimea dâr-e Papar-o dast mi-y-andâz-eK. have-3sg P.-râ hand hab-throw-3sg'Kimea is mocking Papar.'

(49) HV keshidan 'to pull'a. Kimea kâmelan/ barâye ye sâ'at/*ye sâ'ate dast-esh-ro keshid

K. completely/for an hour/within an hour hand-her-râ pulled'Kimea completely/for an hour pulled her hand.'

b. Kimea dâr-e dast-esh-ro mi-kesh-eK. have-3sg hand-her-râ hab-pull-3sg'Kimea is pulling her hand.'

(50) PP + LV be âtash keshidan (to fire pulling) 'to put on fire'a. Kimea xuna-ro kâmelan/* barâye ye sâ'at/ye sâ'ate be âtash keshid

K house-râ completely/for an hour/within an hour to fire pulled'Kimea completely/in an hour put the house on fire.'

b. Kimea dâr-e xuna-ro be âtash mi-kesh-eK. have-3sg house-râ to fire hab-pull-3sg'Kimea is putting the house on fire.'


(51) HV gozashtan 'to pass, cross'a. Kimea kâmelan/*barâye ye sâ'at/ye sâ'ate az xiyâbun gozasht

K. completely/for an hour/within an hour of street passed'Kimea completely/in an hour crossed the street.'

b. Kimea dâr-e az xiyâbun mi-gzar-eK. have-3sg of street hab-pass-3sg'Kimea is crossing the street.'

(52) Particle + LV dar gozashtan (away passing) 'to pass away'a. Kimea kâmelan/*barâye ye sâ'at/?ye sâ'ate dar gozasht

K. completely/for an hour/within an hour away passed'?Kimea died within an hour.'

b. *Kimea dâr-e dar mi-gzar-eK. have-3sg away hab-pass-3sg'Kimea is dying.'

4.3.3 SummaryThe summary of the event structures of the CPrs, some of them presented in this section, is asfollows:

(53) Table TELIC ATELIC

PP + LV Ex:be donyâ âmadan (to world coming) 'to be born'be âtash keshidan (to fire pulling) 'to put on fire'

N + LV Ex:dast xordan (hand colliding) 'to get touched'dâd zadan (scream hitting) 'to yell'gush dâdan (ear giving) 'to listen'dast andâxtan (hand throwing) 'to mock'

Particle + LV Ex:kenâr âmadan (side coming) 'to get along, agree'dar gozashtan (away passing) 'to pass away'A + LV Ex:derâz keshidan (long pulling) 'to take a nap’Eventive Nominal + LV Ex:shekast xordan (defeat colliding) 'to be defeated'shekast dâdan (defeat giving) 'to defeat'farib dâdan (deceit giving) 'to deceive'

As mentioned in the introduction, if the LV is inherently telic, such as shodan 'become', theNV element will not have an effect on the telicity of the whole CPr. The example in (54b)illustrates this fact:

(54) a. xorshid barf-ro âb kardsun snow-râ water made'The sun melted the snow.'

b. barf âb shodsnow water became'The snow melted.'

If the above treatment of telicity is on the right track, the apparent ‘inherent telicity’ of such averb boils down to a selectional restriction: it selects for a predicative small clausecomplement. The telicity of the whole CPr is then still determined by the complement to theLV, not the LV itself. The problem for the purely category-based treatment here, however, isthe fact that above we are assuming that only Adjs and PPs may function as NV predicative


elements. Here, however, a nominal NV element is able to act as a predicate. Apparently,while NV elements of category Adj and P must function as predicates (leading to thegeneralization we present above), NV elements of category N may function as predicates in (avery limited number of) cases, as here.

5 AN EXCEPTION: EVENTIVE NOMINALS

In the chart in (53), there are three cases with NV elements that are nominal and yet in whichthe event structure of the CPr in which they occur is telic – in fact, it is an Accomplishment.Two of them are repeated here:

(55) shekast dâdan (defeat giving) 'to defeat'Kimea dar ye sâ'at/ ye sâ'ate Papar-o shekast dâdK. in one hour/within an hour P.-râ defeat gave'Kimea defeated Papar in one hour.'

(56) shekast xordan (defeat colliding) 'to be defeated'Kimea ye sâ'ate shekast xord K one hour defeat collided'Kimea got defeated in an hour.'

While these seem to be counterexamples to our observation above that NV elements ofcategory N always produce an atelic (activity or semelfactive) reading, in fact, we think theycan be accommodated within the framework if we adopt Harley (2001)’s proposal that thesister-to-v, when it is of category N, functions as an Incremental Theme. Most elements ofcategory N are either themselves unbounded or instantaneous, which leads to thegeneralization above, but if some event-denoting Ns can themselves be telicaccomplishments, we expect them to produce accomplishments in combination with a LV, asthey will function as Incremental Themes. The following phrase structure represents the CPrconsisting of shekast dâdan in (55):

(57) vP

DP v’

Kimea N v

DP N dâdan

Papar-o shekast

The corresponding unaccusative CPr shekast xordan’s underlying structure is presented in(58), created by varying the LV only, of course, as usual. This structure represents otherunaccusatives such as farib xordan 'to be deceived' as well.

(58) vP

N v

DP N xordan

Kimea shekastfarib


In these cases, the nominal NV element itself denotes an event which happens to be anAccomplishment. The event properties of the NV element, then, are inherited by the entireCPr, along the lines proposed by Harley (2001) for bounded and unbounded nominal elementsin English. Compare, for example, the properties of the verb derived from the eventivenominal work (Activity, -bounded N) and knock (Semelfactive, +bounded N). Here, theboundedness of the whole event is therefore expected.

An alternative account of these verbs would involve proposing that they contain acovert PP small clause (in a standard analysis of give/get in languages like English, seePesetsky, 1995, Harley, 1995); however, since Persian shows no overt morphology that wouldconfirm this proposal, and the present paper is attempting to provide the mostmorphosyntactically transparent possible account, we do not consider that possibility here(although see section 7.3).

6 WHAT DETERMINES THE COMPATIBILITY OF AN NV ELEMENT WITH A GIVEN LV?

Although CPr formation is clearly a syntactic process, it is equally clearly not completelyproductive. Certain LVs may not combine with certain NV elements, while others, of course,may. It seems likely to us that at least some of these restrictions themselves reflect generalproperties of the CPr construction, rather than idiosyncratic properties of the lexical itemsinvolved. The following data, for instance, seem to show the effects of the importance of theconcepts of internal vs. external causation, along the lines of Levin & Rappaport’s (1995)proposal concerning the difference between alternating inchoative/unaccusatives (like open)and non-alternating ones (like blush). Consider the example below:

(59) a. Kimea sorx shodKimea red became'Kimea blushed'

b. *Papar Kimea-ro sorx kardPapar Kimea-ro red made*Papar made Kimea blush('Papar fried Kimea')

Because blushing may only be internally caused, sorx ‘red’ may not receive the ‘blush’meaning when it occurs in combination with kardan despite being syntactically unaccusativewhen it occurs in the intransitive form with shod ‘become’, and despite the availability of ashod/kardan alternation for many CPrs illustrated earlier. Similarly, certain NV elementsmay not be combined with the unaccusative xordan, because the events that they denote canonly be caused agentively—they are, in essence, inherently unergative. Accordingly, the ill-formedness of the (b) examples is not syntactic, but semantic.

(60) a. dâd zadan (scream hitting) 'to yell'b. *dâd xordan

(61) a. kâr kardan (work doing) 'to work'b. *kâr shodan


7 SOME CONSEQUENCES AND PREDICTIONS

In this section, we discuss some predictions of our proposal. We start with resultativeconstructions, continue with passives, and finish with a discussion of location/locatum CPrconstructions.

7.1 Resultatives

In a Hale and Keyser-style system, most Accomplishment-denoting verbs are structurallycovert resultatives, resulting from a null causative or inchoative light verb combining with apredicative small clause that denotes the Result (The ice melted). The formation of a trueresultative, with a secondary Result predicate (The ice melted away), in a language likeEnglish, is the product of an exceptional process whereby a verb root like melt is merged inthe place of the causative light v, and the secondary predicate forms the result-denotingpredicative small clause (Harley 2001, Mateu 2002, Folli & Harley forthcoming). Thestructures of each of these two sentences in the present framework are illustrated below:

(62) a. The ice melted b. The ice melted awayvP vP

vBECOME SC vBECOME SC

∅ DP A “melt” DP P

the ice melt the ice away

As is well known, the availability of this sort of ‘manner incorporation’ operation variesparametrically across languages (Talmy 1985); English and the Germanic languages generallyallow it, while Romance languages do not. Whatever the account of the Germanic/Romancevariation, it seems clear that the present analysis predicts that Persian should not allow theformation of such resultatives.

If resultatives result from the ‘merge’ of an ordinarily predicative root in the light verbposition, combined with the insertion of a new resultative predicate low in the structure,resultatives in general should only be possible with NV predicates which are potentiallyverbal in nature. In Persian, of course, change-of-state CPrs are made up of a light verb plus aresultative NV element. Two predictions about resultative formation ensue: (i) Persian shouldnot allow the addition of a secondary predicate to a CPr construction, since the result-predicate slot is already occupied by the NV element; (ii) Persian should not have the optionthat English does, of merging a result-denoting Root in the LV position in order to make roomfor a resultative secondary predicate, because in Persian, the set of light verbs is tightlyconstrained, limited to a few dozen elements at the most.

That is, our analysis thus far predicts that resultatives with complex predicates shouldnot exist in this language since there is no room for complex structure for the secondaryresult-denoting predicate. This prediction is borne out as the following contrast indicates.

(63) a. Kimea xuna-ro rang zadK house-râ color hit'Kimea colored the house.'

b. *Kimea xuna-ro sefid rang zad K house-râ white color hitThe intended meaning: ‘Kimea colored the house white’


Although sefid, ‘white’, cannot be a secondary resultative predicate, the NV element rang'color' can be modified by sefid 'white', as in (64) — recall that NV elements may bestructurally complex.

(64) Kimea be xune rang-e sefid zadK to house color-EZ white hit'Kimea colored the house white' Lit: Kimea hit (on) the house white color

However, (64) does not have the resultive reading Kimea colored the house white. AResultative reading is obtained only by adding a resultative clause, as in (65).

(65) Kimea xuna -ro rang-e sefid kard tâ pro kâmelan sefid shod.K house-râ color-Ez white did till completely white became'Kimea put white color on the house till it became completely white.'

7.2 Passives

Whether or not there is syntactic passive construction in Persian has been highlycontroversial. Given our analysis of Persian complex predicates, it could be argued that thepassive construction is just an instance of CPr, with a past participle serving as its NV element(Karimi, to appear).

(66) ye gol be Papar dâde shod a flower to Paper given was"A flower was given to Papar"

The past participle dâde has adjectival properties. The phrase structure of (66) is provided in(67). The complement of the verbal adjective moves to the subject position.

(67) vP .... AP v

PP A' shod be Papar O A to Papar ye gol dâde a flower given became

This structure is identical to the regular unaccusative CPr consisting of an adjective as the NVelement of LV. Consider the example in (68) and its phrasal structure in (69).

(68) xune xarâb shodhouse destroyed became'The house was destroyed.'

(69) vPAP v

O A shodxune xarâb house destroyed became


Our analysis predicts that there is no passive of atelic verbs with a nominal NV element. Thisis in fact borne out as shown by the following data:

(70) a. hol dâdan (push doing) 'to push'b. *hol dade shodan (push given become)

(71) a. kise keshidan (brush pulling) 'to brush (body)'b. *kise keshide shodan (brush pushed become)

In (70) and (71), we see that the LV which creates the ‘passive’ in combination with adeverbal adjective cannot co-occur with both the deverbal adjective AND a nominal NVelement, which would be necessary in order to form a passive of an atelic CPr with a nominalcomplement. If the deverbal adjective is truly functioning as a NV element in the Persianpassive, this is expected: CPrs can contain only one NV element.

Of course, there is also no unaccusative alternation with these atelic nominal-basedCPrs, where their normal agentive light verb is simply switched for a non-agentive one; this ispresumably for the world-knowledge reasons outlined in section 6 above.

7.3 Location/locatum: Mejerdoomian 2002

Megerdoomian (2002) makes a proposal concerning aspect in CPrs that is in general verycompatible with the view proposed here. She argues, as we have argued above, that the eventstructure of a CPr is the compositional result of the combination of the LV and the NVelement, contra the view of Karimi-Doostan (1997) that it depends entirely on the LV. (Thispoint is also made in Karimi 1997). However, our final conclusion that telicity is presentwhen a predicative SC is present, i.e. with PP and Adj NV predicates, is significantly differentfrom that of Megerdoomian. She argues that it is the presence of a ‘become’ predicate thatensures telicity, whether or not the ‘become’ predicate is overt. Given the complementarydistribution of the inchoative and causative predicates, however (see (24a and b) and (25aand b), as well as (55) and (56) above, for example) we feel that the change in the structure ofthe NV element is the crucial determinant of telicity, rather than the presence of any covertinchoative element in telic causatives.

There is one class of cases discussed by Megerdoomian, however, which at firstglance appear to go against our proposal here: a set of CPrs which can be telic despite havingNV elements which are unambiguously Ns. These are CPrs with meanings like those of theEnglish denominal predicates that Hale and Keyser dub ‘location/locatum’ verbs: shelve, box,saddle, paint, oil, corral, etc. A subset of Megerdoomian’s examples are presented in (72) and(73) below:

(72) a. afsâr zadan ‘to harness’harness hit

b. pâlân zadan ‘to saddle’blanket hit

c. zang zadan ‘to bell’bell hit

(73) a. roqan zadan ‘to oil’oil hit

b. namak zadan ‘to salt’salt hit

c. gard zadan ‘to powder’powder hit


These CPrs have interesting properties which parallel the properties of their Englishcounterparts. According to our proposal above, they should all be atelic, since they are CPrswith nominal NV elements. However, the first group, but not the second are necessarilytelic—exactly like their English counterparts.

This fact about these English predicates was first noted in Harley (1998, 2001), whoargued that for location/locatum verbs, the telicity of the denominal verb was correlated withthe boundedness of the nominal: if the nominal was unbounded (‘mass’) as sand, powder, saltetc., the verb was unbounded; if it was bounded (‘count’), as saddle, bell, bag, etc., the verbwas bounded. Megerdoomian points out that this appears to be true in Persian as well.

We can account for this in the system presented here if we allow for the presence of acovert resultative predicate in the NV element—a preposition—in just this limited class ofcases. Hale and Keyser, recall, propose that location/locatum verbs have the followingstructure, with a covert SC headed by a preposition:

(74) vP

DP v’

Mary v SCJohn

∅ DP PP

the horse P N the soup

∅ saddlesalt

Recall that we asserted above that the boundedness of a CPr with a SC within it wasdetermined by whether or not the SC denoted a scalar state—whether it provided a definiteendpoint (result) or allowed for indefinite increases in the degree of the state. Harley (2001)argued that the boundedness of the state denoted by the covert PPs in locatum verbs dependedon the boundedness of the locatum itself. Megerdoomian has shown that, for the class of CPrswith locatum meanings, this is true in Persian as well. Consequently, we assume that there is acovert prepositional predicate present in these CPrs, providing the locative component: thestructure of pâlân zadan, ‘to saddle’, and namak zadan, ‘to salt’, is given in (75) below.

(75) Structure of pâlân zadan (‘saddle’) and namak zadan (‘salt’)vP

DP v’

Agent SC v

DP PP zadan ‘hit’

Theme N P

pâlân ∅‘blanket’namak ‘salt’


This is the only case in which Persian does not seem to provide a direct morphologicalrealization of every component in Hale and Keyser’s proposed l-syntax. However, the cluesprovided by the aspectual properties of these CPrs, and their locative meaning, combine tosuggest that the analysis proposed by H&K for English should indeed be extended to Persianin these cases as well.

8 CONCLUSIONS

In this paper we have argued that Persian CPrs are syntactically derived from two independentelements: a non-verbal element and a light verb. We have considered in turn the contributionof each element and shown that while the light verb determines the agentivity/causativity, theeventiveness and duration of the CPr, the NV element determines the Aktionsart of eventiveCPrs. These conclusions support a syntax-based approach to verbal composition, as the eventstructure and agentivity of the CPr are direct functions of its individual parts. This division oflabor is not predicted by Lexicalist approaches, which are further faced with the problem ofaccounting for the syntactic independence of the two elements. Persian CPrs directly show thecomplex structure proposed for independent syntactic and semantic reasons in the literaturefor languages like English. Not only do they realize the individual sub-events of verbalstructure as separate morphemes, they realize them as independent syntactic elements, ratherthan as dependent pieces of morphology attached to verbs. Lexicalist approaches, which canargue that complex predicates in many languages should be derived in the lexicon since theyare single phonological words, cannot take that tack with Persian.

REFERENCES

Baker, Mark, Kyle Johnson and Ian Roberts (1989) 'Passive arguments raised.' LinguisticInquiry 20.2:219-251.

Bateni, Mohammad Reza (1989) 'Farsi zabâni aghim?' (Persian a sterile language?). Adine33.

Borer, Hagit. (2002) 'The syntax and semantics of quanitity.' Ms. University of SouthernCalifornia.

Chomsky, Noam (1986) Knowledge of Language. New York: Praeger.Folli, Raffaella and Heidi Harley (2002). 'Waltzing Matilda around the room.' Ms. University

of Arizona, University of CambridgeGhomeshi, Jila (1996) Projection and Inflection: A Study of Persian Phrase Structure. Ph.D

dissertation, University of Toronto.Hale, Ken and Samuel. Jay Keyser (1993) 'On argument structure and the lexical expression

of syntactic relations.' In Hale, Ken and Samuel Jay Keyser (eds.) The View fromBuilding 20. Cambridge, MA: The MIT Press, pp 53-109.

Hale, Ken and Samuel Jay Keyser (2002) Prolegomenon to a Theory of Argument Structure.Cambridge, MA: The MIT Press.

Harley, Heidi (1995) Subjects, Events, and Licensing. Ph.D. dissertation, MIT.Harley, Heidi (1998) 'Denominal verbs and aktionsart.' In Angeliek van Hout, Liina

Pylkkänen, and Heidi Harley (eds.) Papers from the UPenn/MIT Roundtable on theLexicon. Cambridge, MA: MITWPL.

Harley, Heidi (2001) 'Measuring-out and the ontology of verb roots in English.' Talkpresented at the Ben-Gurion University Workshop on Aspect, June 18-20, 2001

Hay, Jennifer, Kennedy, Chris and Levin, Beth (1999) 'Scalar structure underlies telicity?degree achievements?' In T. Mathews and D. Strolovitch (eds.) Proceedings of SALTIX. Ithaca: CLC Publications, pp 127-144.


Karimi, Simin (1997) 'Persian Complex Verbs: Idiomatic or Compositional,' Lexicology 3:273-318.

Karimi, Simin (2003) 'On object positions, specificity, and scrambling in Persian.' In SiminKarimi (ed.) Word Order and Scrambling. Oxford: Blackwell Publishing, pp 91-124.

Karimi, Simin (to appear) A Minimalist Approach to Scrambling: Evidence from Persian,Tucson: University of Arizona.

Karimi-Doostan, Mohamad-Reza (1997) Light Verb Constructions in Persian. PhD.dissertation. University of Essex.

Kratzer, Angelika (1996) 'Severing the external argument from its verb.' In J. Rooryck andA. Zaring, (eds.) Phrase Structure and the Lexicon. Dordrecht: Kluwer

Levin, Beth and Malka Rappaport (1995) Unaccusativity. Cambridge: MIT press.Marantz, Alec (1997) 'No escape from syntax: Don’t try morphological analysis in the

privacy of your own lexicon.' In Dimitriadis, A. and L. Siegel (eds.) University ofPennsylvania Working Papers in Linguistics 4.2: 201-225.

Mateu Fontanaals, Jaume (2002) 'Syntactically-based lexical decomposition: the case of climbrevisited.' In Papers presented at the 26th meeting of the Berkeley Linguistics Society.

Megerdoomian, Karine. (2000a) Beyond Words and Phrases: A Unified Theory of PredicateComposition. Ph.D. dissertation, University of Southern California.

Megerdoomian, Karine. (2002b) 'Aspect in Complex Predicates.' Talk presented at theWorkshop on Complex Predicates, Particles and Subevents, Konstanz University,October 2, 2002.

Mohammad, Jan. and Simin. Karimi (1992) 'Light verbs are taking over: Complex verbs inPersian.' Proceedings of WECOL, pp 195-212.

Pesetsky, David (1995) Zero Syntax: Experiences and Cascades. Cambridge, MA: MIT Press.Talmy, L. (1985). 'Lexicalization patterns: semantic structure in lexical forms.' In Shopen, T.

(eds.) Language typology and syntactic description III: grammatical categories andthe lexicon. Cambridge: Cambridge University Press. 57-149.

Wechsler, Stephen (2001). 'Resultatives under the event-argument homomorphism model oftelicity.' Ms., University of Texas at Austin.

Raffaella Folli Heidi Harley and Simin Karimi

Linguistics Department Department of LinguisticsSidgwick Avenue University of ArizonaUniversity of Cambridge Tucson, AZ 85721Cambridge CB3 9DA USAUnited Kingdom

[email protected]@cam.ac.uk [email protected]

Documents

Determinants of event type in Persian complex predicateshharley/PDFs/HarleyFolliKarimi2003.pdf · Determinants of event type in Persian complex predicates 3 In this paper, we show